Something in the 0.5.0 to 0.5.1 update process causes huge amounts of RAM to be used when loading an existing save.

My existing save went from using around 4GB RAM when loaded, to using over 10GB when first loaded with the 0.5.1 version.

This was a challenge to get working - PC swapping like mad for ages, and factorio crashed due to low memory on the first attempt.

Once I managed to load the game and re-save it, it seems OK, so just something about the upgrade process I think.

Thank you for the heads up. I cannot imagine what could cause that; the memory usage of the mod ought to be a tiny fraction of that of the game.

I have tried to reproduce this bug, but have failed to do so. If you still have the game saved with Bottleneck 0.4.5 that causes the problem, could you send me a link?

It may be another mod, not bottleneck - as something similar occurred just after this when I installed another mod and loaded the save. Bottleneck did show it's 'building from scratch' message this time too, but it's possible another mod is also hooking that event, and causing the issue, but remaining silent about it's activity.

Other mods installed:
- rso-mod_2.3.0
- Bob's mods
- bobassembly_0.14.0.zip
- bobconfig_0.14.0.zip
- bobelectronics_0.14.0.zip
- bobenemies_0.14.0.zip
- bobgreenhouse_0.14.0.zip
- bobinserters_0.14.3.zip
- boblibrary_0.14.3.zip
- boblogistics_0.14.5.zip
- bobmining_0.14.0.zip
- bobmodules_0.14.1.zip
- bobores_0.14.0.zip
- bobplates_0.14.0.zip
- bobpower_0.14.0.zip
- bobrevamp_0.14.1.zip
- bobtech_0.14.0.zip
- bobvehicleequipment_0.14.2.zip
- bobwarfare_0.14.2.zip
- Auto_Deploy_Destroyers_0.1.5
- RadarGridGuide_0.1.3
- add-loader_0.0.8
- compoundsplitters_0.2.8
- Inventory Sensor_1.1.5
- nixie-tubes_0.3.10
- personal-roboport-switch_0.1.4
- ScoreExtended_1.0.22
- UsefulMapColors_1.1.5
- Warehousing_0.0.11

Here's the save game and set of mods that caused me the RAM problems.

http://www.jonatkins.com/factorio/bugs/mod-upgrade-ram-0.14/

I will take a look into this tonight also as I was not able to reproduce when testing. but I didn't test the .0.4.5 to 0.5.1. My tests were .0.5.0 to 0.5.1 and show a huge speedup in world load times and a slight decrease in script times from about .52/t to .48/t average on a heavily modded base.

I can confirm the bug. I have moved the description to github.
https://github.com/troelsbjerre/Bottleneck/issues/17
I didn't have time to look at it properly, but there is definitely some weird interaction going on. The upgrade of Bottleneck from 0.5.0 to 0.5.1 is needed to make the bug show up; keeping it at 0.5.0 or removing it avoids the memory spike (my laptop didn't recover after it started swapping, so I can't tell whether it actually would have gotten through the upgrade).

@jonatkins Check the issue for the latest news on this
https://github.com/troelsbjerre/Bottleneck/issues/17